Order of SignatureLength, Signature in StandardTrailer

[vanrein]

Had to get into QuickFIX/Go, but it's growing on me. Thanks!

I'm building a FIX application using OpenPGP-signed FIXT messages.
This revealed what I found to be a bug in the order of StandardTrailer fields, as coded in

quickfix/message.go

Lines 87 to 97 in 2ed31c3

	// In the trailer, CheckSum (tag 10) must be last.
	func trailerFieldOrdering(i, j Tag) bool {
	switch {
	case i == tagCheckSum:
	return false
	case j == tagCheckSum:
	return true
	}
	return i < j
	}

This implicitly places Signature(89) before SignatureLength(93), instead of the intended order,

• SignatureLength(93) if used at all
• Signature(89) if used at all
• Checksum(10)

Placing SignatureLength after Signature impairs customary front-to-back parsing.

Since these are the only three fields in the StandardTrailer component, reverse ordering of tag values would fix the problem (until a fourth field is added perhaps).

Minimal code to demonstrate this bug:

package main

import "fmt"
import qfapi "github.com/quickfixgo/quickfix"

func main () {
        msg := qfapi.NewMessage();
        msg.Trailer.SetField (93, qfapi.FIXInt (5))
        msg.Trailer.SetField (89, qfapi.FIXString ("VROEM"))
        fmt.Print (msg, "\r\n")
}

It does not help to manually map to bytes to have msg.cook() invoked.
The order of the msg.Trailer.SetField() calls also has no impact.
All those non-workarounds make sense, having seen the trailerFieldOrdering() implementation.

Let me know if you'd like a PR, but I'm new to QuickFIX and don't default to it.

[vanrein]

FYI, I tested the reverse tag ordering and it works-for-me, and probably for anyone else too:
https://gitlab.com/arpa2/quickfixgo/-/commit/a6c125ece626a74a06b1a51b49924627fa3de68c